Dell PowerScale F710 vs F910
Dell PowerScale F710
Dell PowerScale F910
The Dell PowerScale F710 and F910 are both all-flash, all-NVMe scale-out NAS nodes running the OneFS operating system, and both sit at the high end of the PowerScale all-flash line. They share the same single-namespace architecture, the same 3-node cluster minimum and 252-node maximum, DDR5 memory, and the same OneFS data services, including inline compression and deduplication. The real difference is the chassis and what it enables: the F710 is a 1U node built on the PowerEdge R660 for performance density, while the F910 is a 2U node built on the PowerEdge R760 that packs far more NVMe drives per node for maximum all-flash capacity. The decision comes down to whether you are optimizing for performance-per-rack-unit and node-level parallelism, or for the highest flash capacity in the fewest nodes.
Side by side
| Dell PowerScale F710 | Dell PowerScale F910 | |
|---|---|---|
| Positioning | 1U performance node, peak throughput and low latency with strong performance density per rack unit | 2U maximum-capacity node, the densest all-flash node in the F-series for the largest flash data sets |
| Form factor & platform | 1U, based on Dell PowerEdge R660 with DDR5 memory | 2U, based on Dell PowerEdge R760 with DDR5 memory |
| NVMe drives per node | Up to 10 all-flash NVMe SSDs per node | Up to 24 all-flash NVMe SSDs per node |
| Capacity per node | High raw capacity per node, roughly 38 TB to 307 TB depending on drive size | Substantially higher raw capacity per node, the highest in the all-flash line, driven by up to 24 drives per chassis |
| Cluster raw capacity | Scales to approximately 77 PB raw per cluster | Scales to meaningfully higher raw capacity per cluster, the largest all-flash ceiling in the F-series |
| Memory per node | 512 GB DRAM per node for large cache and fast metadata handling | Larger per-node DRAM ceiling than the F710 to feed a denser, higher-capacity node |
| Performance profile | More nodes per petabyte means more aggregate CPU and network per TB, excellent for streaming throughput and latency-sensitive concurrency | Very high per-node and aggregate throughput, with fewer nodes carrying more capacity for large sequential and mixed workloads |
| Density & consolidation | Compact 1U footprint; reach large capacity by scaling node count, which also scales performance | Fewer, denser 2U nodes for a given capacity, reducing node count, rack units, and cluster management overhead at scale |
| Best-fit workloads | AI/ML training and inference, high-performance analytics, HPC scratch, media and entertainment pipelines | Large GenAI and analytics data lakes, high-resolution media libraries, and consolidation of big all-flash data sets into fewer nodes |
Shop these now
Live configurations from our catalog with partner pricing. Add to your cart to request a firm quote, or build a full BOM.
Dell PowerScale F710
Dell PowerScale F910
Choose Dell PowerScale F710 when
Reach for the F710 when performance density is the priority and you want maximum compute and network resources per terabyte. Its 1U chassis, up to 10 NVMe drives, and 512 GB of memory make it the node for latency-sensitive AI/ML training and inference, high-performance analytics, HPC scratch space, and media pipelines that push storage hard under heavy concurrency. Because you add capacity by adding nodes, cluster performance scales right alongside capacity, which suits workloads that are bottlenecked on throughput rather than raw space. It is also the tighter footprint when rack-unit-level performance, not total capacity, is what you are buying. If your pipelines need the most parallelism per petabyte, the F710 is the fit. Send the workload profile to /quote and Uniqcli will size the node count.
Choose Dell PowerScale F910 when
The F910 is the right call when you need the most all-flash capacity in the fewest nodes. Its 2U chassis holds up to 24 NVMe drives, giving the highest raw capacity per node in the F-series and a larger per-node memory ceiling to match. That makes it ideal for large GenAI and analytics data lakes, high-resolution media and rendering libraries, and consolidation projects where you want to collapse many smaller nodes into fewer dense ones to cut rack units, power, and cluster management overhead. You still get full NVMe performance and the same OneFS data services, just optimized for capacity density rather than performance-per-rack-unit. When the deal is about scaling flash capacity efficiently at petabyte scale, lead with the F910. Drop the target capacity into a /bom and Uniqcli will model the node layout.
Both nodes run the same OneFS and can coexist in a single cluster and namespace, so this is a sizing decision, not a platform choice. Pick the F710 when performance density drives the deal: latency-sensitive AI/ML, high-performance analytics, HPC, and media pipelines where more nodes per petabyte means more throughput and parallelism per terabyte. Step up to the F910 when capacity density is the goal, using its 2U, 24-drive chassis to pack the most all-flash capacity into the fewest nodes for large data lakes, media libraries, and consolidation at petabyte scale. Because OneFS mixes node types, many customers pair them, with F710 nodes serving the hottest, most performance-critical data and F910 nodes carrying the bulk all-flash capacity. For federal and research buyers, both are available through TAA-compliant, contract-aligned procurement paths. Uniqcli sells both and can right-size the node mix to your workload, budget, and growth curve; start a /quote or build a /bom to get a configuration modeled.
Talk to a specialistFrequently asked
Can the F710 and F910 run in the same PowerScale cluster?
Yes. Both are PowerScale all-flash nodes running OneFS, which presents a single namespace and single filesystem across mixed node types. You can combine F710 and F910 nodes in one cluster, for example using the F710 for hot, performance-critical data and the F910 for high-capacity flash, and let OneFS handle data placement and tiering. A cluster needs a minimum of three nodes and supports up to 252.
What is the main hardware difference between the F710 and F910?
Chassis and drive count. The F710 is a 1U node on the PowerEdge R660 with up to 10 NVMe drives and 512 GB of memory. The F910 is a 2U node on the PowerEdge R760 with up to 24 NVMe drives and a larger memory ceiling, which gives it substantially higher raw capacity per node and the highest capacity density in the all-flash line. Both use DDR5 and deliver full NVMe performance.
Which one is better for AI workloads?
Both handle AI well, but they optimize differently. The F710 favors performance density, delivering more aggregate throughput and parallelism per terabyte, which suits latency-sensitive training and inference. The F910 favors capacity density, making it strong for large GenAI and analytics data lakes where you want maximum flash capacity in fewer nodes. If storage throughput is your bottleneck, lean F710; if the data set size is the challenge, lean F910. Uniqcli can help model both against your pipeline.
Is the F910 just a bigger F710?
Not exactly. The F910 is a purpose-built 2U capacity-dense node, not simply a larger F710. It more than doubles the drives per node and adds memory headroom to feed that capacity, so it changes the node-count-to-capacity ratio of a cluster. The F710 keeps a compact 1U footprint that scales performance as you add nodes. They are complementary points in the same all-flash line rather than one replacing the other.
Are these available on federal contracts?
Yes. PowerScale F710 and F910 nodes can be procured through TAA-compliant, contract-aligned paths that federal, defense, and research buyers use, and Uniqcli supports GSA, NASA SEWP V, and GPC purchasing workflows. Send your requirement to /quote or build a /bom and Uniqcli will align the configuration to the right vehicle.
Build your Dell bill of materials.
Send us the requirement, the project, or an existing quote to beat. We come back with a validated, TAA-compliant Dell configuration and a real price, often below list.
[email protected] · Chicago, IL